Water Concentration in Single‐Crystal (Al,Fe)‐Bearing

Home , Ferropericlase

RESEARCH LETTER Water Concentration in Single‐Crystal (Al,Fe)‐Bearing 10.1029/2019GL084630 Bridgmanite Grown From the Hydrous Melt: Implications Key Points: • High‐quality, inclusion‐free for Dehydration Melting at the Topmost Lower Mantle bridgmanite single crystals Suyu Fu1 , Jing Yang1,2 , Shun‐ichiro Karato3 , Alexander Vasiliev4,5,6, (Mg Fe3+ Fe2+ Al Si O ) 0.88 0.065 0.035 0.14 0.90 3 4 6,7,8 6,7 were synthesized and characterized Mikhail Yu. Presniakov , Alexander G. Gavrilliuk , Anna G. Ivanova , • The crystals contain ~1,020(±70) Erik H. Hauri9,10 , Takuo Okuchi11 , Narangoo Purevjav11 , and Jung‐Fu Lin1 ppm wt water using NanoSIMS and show pronounced OH‐stretching 1Department of Geological Sciences, Jackson School of Geosciences, The University of Texas at Austin, Austin, TX, USA, ‐1 bands at ~3230 and ~3460 cm in 2Geophysical Laboratory, Carnegie Institution of Washington, Washington, DC, USA, 3Department of Geology and FTIR spectra 4 • Dehydration melting at the topmost Geophysics, Yale University, New Haven, CT, USA, National Research Center, Kurchatov Institute, Moscow, Russia, 5 6 lower mantle can occur when water Moscow Institute of Physics and Technology, Moscow, Russia, Shubnikov Institute of Crystallography of Federal content exceeds ~0.1 wt% solubility Scientific Research Center Crystallography and Photonics, Russian Academy of Sciences, Moscow, Russia, 7Institute for limit Nuclear Research, Russian Academy of Science, Moscow, Russia, 8REC Functional Nanomaterials, Immanuel Kant Baltic Federal University, Kaliningrad, Russia, 9Department of Terrestrial Magnetism, Carnegie Institution of Washington, Supporting Information: Washington, DC, USA, 10Deceased, 11Institute for Planetary Materials, Okayama University, Misasa, Japan • Supporting Information S1 • Data Set S1 3+ 2+ Abstract High‐quality single‐crystals of (Al,Fe)‐bearing bridgmanite, Mg0.88 Fe 0.065Fe 0.035Al0.14Si0.90O3, of hundreds of micrometer size were synthesizedat24GPaand1800°CinaKawai‐type apparatus from the Correspondence to: J.‐F. Lin and S.‐i. Karato, starting hydrous melt containing ~6.7 wt% water. Analyses of synthesized bridgmanite using petrographic [email protected]; microscopy, scanning electron microscopy, and transmission electron microscopy show that the crystals are ‐ shun [email protected] chemically homogeneous and inclusion free in micrometer‐ to nanometer‐spatial resolutions. Nanosecondary ion mass spectrometry (NanoSIMS) analyses on selected platelets show ~1,020(±70) ppm wt water Citation: (hydrogen). The high water concentration in the structure of bridgmanite was further confirmed using ‐ Fu, S., Yang, J., Karato, S. i., Vasiliev, polarized and unpolarized Fourier‐transform infrared spectroscopy (FTIR) analyses with two pronounced A., Presniakov, M. Y., Gavrilliuk, A. G., ‐ −1 ‐ et al. (2019). Water concentration in OH stretching bands at ~3,230 and ~3,460 cm .Ourresultsindicatethatlowermantle bridgmanite can single‐crystal (Al,Fe)‐bearing accommodate relatively high amount of water. Therefore, dehydration melting at the topmost lower bridgmanite grown from the hydrous mantle by downward flow of transition zone materials would require water content exceeding melt: Implications for dehydration melting at the topmost lower mantle. ~0.1 wt%. Geophysical Research Letters, 46. https://doi.org/10.1029/2019GL084630 Plain Language Summary Water cycle between surface oceans and Earth's deep interior is a key to understanding the evolution and physical/chemical states of the planet. Early studies show that Received 19 JUL 2019 major transition zone minerals, wadsleyite, and ringwoodite, could accommodate abundant water (1–3 Accepted 12 AUG 2019 wt%), in the form of lattice‐bonded hydrogen atoms, in their crystal structures. However, water solubility in Accepted article online 16 AUG 2019 lower‐mantle bridgmanite, the most abundant mineral in the most volumetric layer of the planet, has remained poorly understood. The scientific challenge here was largely due to difficulties in making large‐ sized high‐quality single‐crystals of bridgmanite for reliable characterizations of its water concentration. Here we synthesized single‐crystal bridgmanite of a few hundred micrometers in diameter, which are examined to be inclusion and precipitate free and thus can be used for reliable water concentration measurements using NanoSIMS analyses. Unpolarized and polarized FTIR analyses are used to identify characteristic OH‐stretching bands. Our results show that (Al,Fe)‐bearing bridgmanite could contain as high as 1,020(±70) ppm wt water. This high water concentration in bridgmanite has implications for our understanding of how melting can occur deep in the mantle below the transition zone.

1. Introduction Water (hydrogen) can be dissolved into the structures of most mantle minerals including those that nominally do not have hydrogen in their chemical formulae such as olivine and its high‐pressure

polymorphs, wadsleyite, and ringwoodite, (Mg,Fe)2SiO4, called as nominally anhydrous minerals (NAMs; e.g., Bolfan‐Casanova, 2005; Karato, 2015; Peslier et al., 2017). Water in NAMs can have signiﬁcant effects

©2019. American Geophysical Union. on a variety of properties, including melting relationships (Inoue, 1994; Kawamoto, 2004), rheological All Rights Reserved. properties (Karato & Jung, 2003; Mei & Kohlstedt, 2000), and electrical conductivity (Karato & Wang,

FU ET AL. 1 Geophysical Research Letters 10.1029/2019GL084630

2013). For instance, early studies indicate that partial or dehydration melting would occur when water‐rich materials in the transition zone such as wadsleyite or ringwoodite are transported to either upper or lower mantle regions where water solubility is low (e.g., Bercovici & Karato, 2003; Liu et al., 2016; Schmandt et al., 2014; Tauzin et al., 2010). When melting occurs, a majority of water would strongly partition into the melt. Since the melt is generally mobile, melting near the transition zone regions could lead to large‐ scale transport of water in the Earth's mantle. Therefore, understanding the behavior of hydrogen in mantle minerals is important for our knowledge of the dynamics and evolution of the Earth. A major goal of previous work has been to determine the solubility of water in mantle NAMs at various pressures, temperatures, and oxidation states. There has been a reasonable agreement on research results for upper mantle and transition zone minerals (e.g., Karato, 2015; Ohtani, 2015; Peslier et al., 2017). However, water solubility in the lower‐mantle minerals, particularly for bridgmanite, has been poorly constrained. Previous studies suggested that bridgmanite can contain either low (a few ppm wt, near or less than the detection limit) or high (~2,000 ppm wt) water concentrations (e.g., Bolfan‐Casanova et al., 2003; Inoue et al., 2010; Litasov et al., 2003; Meade et al., 1994; Murakami et al., 2002; Panero et al., 2015). Consequently, the conditions at which melting would occur at the topmost lower mantle by downward ﬂow of transition zone minerals are not well constrained. The cause of large variations in reported water concentrations in bridgmanite is not well known but includes differences in synthesis conditions at high pressure‐temperature (P‐T), compositions and water contents of starting materials, and quenching rates; the use of different analytical techniques and their calibrations; total Fe, Al as well as Fe3+ contents in bridgmanite; and possible contaminations from inclusions of hydrous phases and/or precipitates (e.g., Kaminsky, 2018). In the next two sections, we will present a review to identify possible causes for different results on water solubility in bridgmanite from previous studies. The potential role of inclusions on reported results is emphasized in the discussion below.

1.1. The Use of Analytical Techniques In order to characterize water in mantle minerals, Fourier‐transform infrared spectroscopy (FTIR) and secondary ion mass spectrometry (SIMS) have been commonly used (e.g., Pearson et al., 2014; Sambridge et al., 2008). FTIR can be used to detect characteristic structural OH absorption bands typically in the range of − 3,000–3,800 cm 1. The main advantage of FTIR is that the nature of hydrogen‐related species can be inferred from the frequencies of absorption. Consequently, absorption by inclusions and/or surface contaminants could be distinguished from absorption by hydrogen dissolved in crystallographic sites of the candidate crystals. To estimate hydrogen content from IR absorption, one needs reliable calibrations of absorption coefﬁ- cients in speciﬁc crystallographic orientations relative to the polarization direction of the IR beam (Asimow et al., 2006; Aubaud et al., 2007; Balan et al., 2008; Paterson, 1982). However, this calibration standard has not been established yet for bridgmanite. In contrast, SIMS can be used to measure bulk water contents (e.g., Hauri et al., 2011; Kumamoto et al., 2017; Mosenfelder et al., 2011). Nevertheless, in SIMS analyses one cannot distinguish hydrogen in inclusions (e.g., glasses and secondary phases), precipitates, and surface contaminants (e.g., hydroxyls) from hydrogen in crystalline lattices of the sample, if one of these sources is present.

1.2. The Role of Inclusions Experiments on water partitioning between ringwoodite and bridgmanite using multianvil apparatus found that water preferentially partitions into ringwoodite, and ~1–2 ppm wt water was reported to be barely

detectable in MgSiO3 or Fe‐bearing bridgmanite using FTIR analysis (Bolfan‐Casanova et al., 2000; Bolfan‐Casanova et al., 2003). The synthesized single‐crystal bridgmanite was in coexistence with hydrogen‐rich ringwoodite and contained numerous tiny inclusions of super hydrous phase B and brucite (Bolfan‐Casanova et al., 2000; Bolfan‐Casanova et al., 2003; Inoue et al., 2010). Furthermore, recent experiments suggested that bridgmanite synthesized from both hydrous and dry enstatite in laser‐heated diamond anvil cells contained less than 220 ppm wt water (Panero et al., 2015). The synthesized bridgmanite coexisted with polycrystalline stishovite, Pt metal (as a laser absorber), and melt (Panero et al., 2015). We should note that FTIR technique does not have the spatial resolution to distinguish water in bridgmanite from these coexisting phases. Water may have strongly partitioned into stishovite and/or Pt metal (Litasov et al., 2007; Schwager et al., 2004; Spektor et al., 2011), lowering the amount of water in bridgmanite below what

FU ET AL. 2 Geophysical Research Letters 10.1029/2019GL084630

is possible for bridgmanite coexisting with hydrous melt. In contrast to these studies, high water concentration, as much as 0.2 wt%, was reported in (Al,Fe)‐bearing bridgmanite using FTIR and SIMS analyses (Murakami et al., 2002). However, Schmandt et al. (2014) observed the formation of intergranular melt around bridgmanite grown from hydrous ringwoodite, and the intergranular melt displayed the same FTIR spectra as observed by Murakami et al. (2002). This indicated that the high water solubility of ~0.2 wt% reported by Murakami et al. (2002) was likely due to the presence of intergranular hydrous melt inclusions between grain boundaries. We note that the IR spectra of ferropericlase and bridgmanite collected by − Murakami et al. (2002) are almost identical with a broad band at 3,200–3,600 cm 1, suggesting possible contaminations from hydrous melt inclusions. On the other hand, Litasov et al. (2003) reported 100 ppm wt

water in MgSiO3 bridgmanite from FTIR measurements. Their studies further indicated that Al‐bearing bridgmanite exhibits signiﬁcantly higher water solubility as a result of Al substitution induced vacancies

in the lattice: bridgmanite containing 4–7 wt% Al2O3 can contain as high as 1,000–1,500 ppm wt water (Litasov et al., 2003). Furthermore, (Al,Fe)‐bearing bridgmanite in various mantle rock types can contain different amounts of water (Litasov et al., 2003): the water concentration was suggested to be 40–110 ppm wt in

the MORB system (Al2O3 =13–17 wt%, Fe/(Fe + Mg) = 0.39–0.42) and 1,400–1,800 ppm wt in the peridotite system (Al2O3 =5–6 wt%, Fe/(Fe + Mg) = 0.1–0.12). Litasov et al. (2003) also reported hydrous mineral inclusions in thick samples. Some sharp absorption peaks overlaid the broad bands in their IR spectra, which might be indicative of contaminants. In summary, we believe the presence of hydrogen‐rich inclusions is the most serious cause for the discre- pancy of previous results on water concentration in bridgmanite. The origins of inclusions are not well understood but include the fast crystal growth or formation of precipitates during temperature quenching. We should note that if a bridgmanite sample contains hydrogen‐rich inclusions (melt, super hydrous phase B, etc.), the water concentration in inclusion‐free regions is controlled by water partitioning between the clean and inclusion‐rich regions. Because hydrogen preferentially partitions into hydrogen‐rich phases, the reported water concentration from such a sample will be strongly biased. Consequently, the results in early bridgmanite samples with inclusions cannot truly reflect its water solubility under relevant lower‐ mantle P‐T and compositional conditions where hydrogen‐rich phases do not exist. In order to obtain robust water solubility in bridgmanite, it is essential to synthesize high‐quality bridgmanite crystals without inclusions/contaminants and characterize them carefully using multiple analytical techniques. 3+ 2+ Here we synthesized large‐size single‐crystal (Al,Fe)‐bearing bridgmanite, Mg0.88Fe 0.065Fe 0.035Al0.14Si0.90O3, using a Kawai‐type apparatus at 24 GPa and 1800 °C with a heating duration of 20 hr. In order to characterize its quality, including (1) homogeneity of major element chemistry; (2) amount of Fe3+; (3) presence or absence of inclusions and/or precipitates, we used electron microprobe analysis (EPMA), scanning electron microscopy‐energy dispersive spectroscopy (SEM‐EDS) mapping, transmission electron microscopy (TEM) including high‐resolution TEM (HRTEM) images, X‐ray diffraction (XRD) and micro‐Raman spectroscopy analyses. The crystals are examined to be chemically homogeneous and inclusion free in micrometer‐ to nanometer‐spatial resolution. NanoSIMS measurements show the (Al,Fe)‐bearing bridgmanite contains ~1,020(±70) ppm wt water. FTIR analyses show pronounced OH‐stretching bands that indicate the presence of significant amounts of lattice‐ bonded water in bridgmanite. The high‐water concentration in bridgmanite is further applied to better understand dehydration melting by downward flow of transition zone materials at the topmost lower mantle.

2. Experiments Single‐crystal (Al,Fe)‐bearing bridgmanite with run number 5K2667 was synthesized at 24 GPa and 1800 °C using a Kawai‐type cell at the Institute for Planetary Materials at Okayama University. The cell was equipped with Fujilloy F08 tungsten carbide anvils with 6‐mm truncation edge length and sintered MgO‐

Cr2O3 pressure media with an octahedral edge length of 14 mm (Okuchi et al., 2015). Pressure calibration for this assembly was well established in previous studies (Shatskiy et al., 2009; Shatskiy et al., 2011).

Starting sample mixture, including magnesium hydroxide [Mg (OH)2], enstatite [MgSiO3], aluminum oxide 57 [Al2O3], and iron oxide [ FeO] in weight percentages of 21.6, 58.8, 11.8, and 7.8 wt%, respectively, were cold sealed into a Pt capsule, placed into the experimental assembly, and then compressed to ~24 GPa. The

FU ET AL. 3 Geophysical Research Letters 10.1029/2019GL084630

capsule was then heated up to ~1800 °C by supplying an electric power of 580 watts to the LaCrO3 heater. Under these P‐T conditions, the sample mixture melted (Kawamoto, 2004) and bridgmanite crystals grew from the hydrous melt that initially contained ~6.7 wt% water (see supporting information). Preestablished power‐temperature relations from literature (Okuchi et al., 2015) were used to control the temperature in the sample capsule within approximately ±6% uncertainty. We used a long heating duration of ~20 hr to allow slow growth and formation of large and chemically homogenous bridgmanite crystals (Okuchi et al., 2015). This slow growth method is distinct from earlier studies using fast growth with large temperature gradients (Shatskiy et al., 2009; Shatskiy et al., 2011) or short‐heating duration method (e.g., Meade et al., 1994; Murakami et al., 2002), which could result in hydrous inclusions in bridgmanite crystals. At the end of the synthesis, the electric power was shut down to quench the sample capsule to room temperature. In the decompressed and recovered sample capsule, large subhedral bridgmanite crystals with maximum grain sizes of ~400 μm in length were embedded in ﬁne ground powder matrix (Figure 1a). The bridgmanite crystals display a light brownish color and can be easily extracted out from the surrounding matrix (Figure 1b).

Bridgmanite crystals and the matrix were initially analyzed with micro‐Raman spectroscopy (see experimental details in supporting information). Collected Raman spectra in our bridgmanite crystals show sharp −1 peaks at ~250, ~280, ~377, and ~496 cm that are consistent with reported vibrational bands in MgSiO3 bridgmanite (Figures 1c and 1d; Durben & Wolf, 1992; Hemley et al., 1989). Some broad peaks at ~574, − ~709, ~734, and ~881 cm 1 are probably due to the Fe and Al chemical substitutions (Litasov et al., 2003). On the other hand, the quenched matrix (Figures 1c and 1d) displays sharp Raman peaks at ~213, ~329, − and ~1,094 cm 1 that could be attributed to super hydrous phase B (Hofmeister et al., 1999; Ohtani et al., − 2001), while the peak at ~3,640 cm 1 is characteristic of brucite phase (Bolfan‐Casanova et al., 2003; Duffy et al., 1995). Overall, micro‐Raman results show that the quenched ground matrix contains predomi- nantly brucite and super hydrous phase B. Additionally, XRD analyses show these two phases in the ground matrix make up approximately 50% and 50%, respectively (Figure S1). We should note that the super hydrous phase B and brucite in the powder matrix were formed from residual melt during high‐temperature quenching. Our synthesized bridgmanite crystals do not show any signatures of hydrous inclusions within the spectral and spatial resolution of micro‐Raman spectroscopy.

The synthesized bridgmanite was further characterized using EPMA, Mössbauer spectroscopy, SEM‐EDS mapping, synchrotron XRD, electron diffraction, and TEM analyses (see experimental details in Text S1 in supporting information). Combination of Mössbauer and EPMA results shows the crystals have an average 3+ 2+ chemical formula of Mg0.88Fe 0.065Fe 0.035Al0.14Si0.90O3, on the basis of three oxygen atoms within experimental uncertainties (Figure S2 and Table S1). Further SEM‐EDS mapping of Fe, Al, Si, Mg, and O shows a chemically homogeneous single‐crystal sample in micrometer‐scale resolution (Figure S3a). Analyses of synchrotron XRD patterns at ambient conditions show the sample crystallizes in the Pbnm space group with refined unit cell parameters of a = 4.7875(3) Å, b = 4.9423(2) Å, c = 6.9205(6) (Figures 2a and 2b). Average full width at half maximum (FWHM) of integrated peaks is ~0.040°, indicating high‐quality crystallinity of synthesized bridgmanite. Additionally, collected selected area electron diffraction patterns on crystals in [221] and [121] zone axes confirmed its high‐quality crystallinity without extra reflections (Figure 2c; Kojitani et al., 2007). We note that the refined unit cell parameters from selected area electron diffraction patterns are consistent with synchrotron XRD measurements within 1% accuracy. Further observations on a double‐side polished crystal of ~30‐μm thick with a petrographic microscope confirmed no apparent inclusions or twining of the sample (Figure S3b). In order to examine the quality of crystals in nanometer‐ scale resolution, HRTEM images were collected and simulated using EMS package (Stadelmann, 1987) over a wide range of thickness‐defocus conditions to resolve the best fits (Figure S4). HRTEM results show that bridgmanite crystals have no visible inclusions or precipitates in nanometer‐scale resolution but exhibit high density of structural defects (Figure 2d), which were probably due to Fe/Al substitution in the lattice and/or stress from high pressure‐temperature quenching.

For FTIR and NanoSIMS measurements, we selected several pieces of big crystals, ~100 μm in length, with clear subhedral shapes, and double‐side polished them to a thickness of 20–25 μm using 3M diamond ﬁlms. The thickness of each platelet was measured under a scaled stereo microscope with an accuracy of ±2 μm. To reliably determine their crystallographic orientations, synchrotron single‐crystal XRD was conducted on

FU ET AL. 4 Geophysical Research Letters 10.1029/2019GL084630

Figure 1. Sample synthesis and micro‐Raman characterizations of quenched products in the Pt capsule. (a) Schematic ﬁgure of the recovered capsule with large subhedral bridgmanite crystals (red) and ground matrix (light green). Large‐size single‐crystal bridgmanite grew from the hydrous melt at high P‐T conditions over a 20‐hr heating duration with a slow growth rate. After high‐temperature quenching, the residual melt turned into ground matrix containing super hydrous phase B and brucite, which surrounded the bridgmanite crystals. (b) Optical stereo microscope images of the recovered and polished bridgmanite crystal and quenched ground matrix using both transmitted and reﬂected lights. Selected areas of the bridgmanite crystal and ground matrix for Raman measurements are labeled as dots A, B, and C. (c and d) Collected − Raman spectra of selected areas at micro‐Raman shift ranges of 200–1,300 and 3,200–4,000 cm 1, respectively. Vibrational bands of bridgmanite are labeled with respective Raman frequency shifts (black). Raman spectra from ground matrix show − sharp vibrational bands (labled with blue), including the distinct OH‐stretching at 3,640 cm 1.

platelets with rotations by ±16° about the vertical axis of the sample stage. Two platelets with crystallographic orientations of (−0.50, 0.05, −0.86) and (0.16, 0.13, −0.98), where the numbers correspond to the direction cosines in the crystallographic coordinates, denoted as P1 and P2 platelets, respectively, were used for measurements (Figure S5). We note that these two platelets were chosen due to the large surface areas after polishing and a dihedral angle of ~53.6° between them that is sufﬁcient to display the orientation effects, if any are present. Unpolarized and polarized FTIR experiments were conducted on selected platelets before NanoSIMS measurements, because FTIR is nondestructive to the sample while NanoSIMS is destructive. FTIR experiments were performed at Yale University using a Varian 600 UMA FTIR spectrometer. The crystals were cleaned multiple times using acetone to remove potential contaminants from epoxy, and potential acetone left on the crystal surface were then removed using ethanol multiple times. To minimize possible noise, dry air was ﬂushed into the FTIR spectrometer installation room, which reduced the detectable water content down to the lowest level of ~1 ppm wt (Dai & Karato, 2009). Unpolarized infrared spectra were taken on the sample − − with 2,048 scans from 550 to 8,000 cm 1 with a spectral resolution of 2 cm 1 (Figure 3a). An aperture of 50 × 50 μm was used for the data collection. The obtained spectra were corrected by a linear baseline between − 3,000 and 3,780 cm 1, which were used to calculate the absorbance. Additionally, polarized FTIR spectra were collected from the P1 platelet by rotating about its normal axis at an azimuthal angle of 10° apart (Figures 3b and S6). After FTIR experiments, identical crystal platelets were sent to Department of Terrestrial Magnetism, Carnegie Institution for Science for NanoSIMS analyses using a Cameca NanoSIMS 50L. Three basaltic glass standards with known volatile concentrations were analyzed for calibrations before each analytical session (Lloyd et al., 2014). Presputtering was conducted on each spot using a 10 × 10 μm raster to remove potential

FU ET AL. 5 Geophysical Research Letters 10.1029/2019GL084630

Figure 2. X‐ray diffraction and transmission electron microscopy analyses of single‐crystal bridgmanite. (a) A representative XRD image of the crystal with the orientation matrix determined using GSE‐ADA/RSV software (Dera et al., 2013). (b) Integrated XRD pattern with main diffraction peaks labeled with identiﬁed miller indices (hkl). Integrated XRD peaks show an average FWHM of ~0.040°. The wavelength of the incident X‐ray beam was 0.3344 Å. (c) Collected electron diffraction patterns in [221] (left) and [121] (right) zone axes. Miller indices (hkl) were labeled close to representative diffraction spots. (d) Fourier‐ﬁltered high‐resolution transmission electron microscopy image of the single‐crystal bridgmanite in the [010] zone axis.

surface contaminants. After the presputter, the area of the raster was reduced to 7.5 × 7.5 μm for NanoSIMS analyses. Following the procedure in literatures (Hauri et al., 2011; Lloyd et al., 2014), three to four spots were selected on each platelet for NanoSIMS measurements on hydrogen and other volatile elements, C, S, F, and Cl (Figures 3c and 3d).

3. Results and Discussion In collected unpolarized spectra (Figure 3a), both platelets display two pronounced absorption bands − − centered at ~3,210 and ~3,465 cm 1 for P1 platelet and at ~3,230 and ~3,475 cm 1 for P2 platelet with −1 an average FWHM of ~120 cm . Comparing with literature reports on MgSiO3,Al‐bearing and (Al,Fe)‐ − bearing bridgmanite (Litasov et al., 2003), we could notice that a broad peak centered at 3,400 cm 1 − with FWHM of ~250 cm 1 was typically observed in FTIR spectra (Figure S7 and Table S2). An addi- − tional weak peak centered at ~3,100–3,200 cm 1 also existed in (Al,Fe)‐bearing bridgmanite in both MORB and peridotite systems (Litasov et al., 2003). Thus, the two absorption bands centered at approxi- − mately 3,210–3,230 and 3,465–3,475 cm 1 observed in our bridgmanite crystals could be assigned to be OH‐stretching bands. Polarized FTIR spectra of P1 platelet show similar peak positions but signiﬁcant − intensity variations, especially for the band at ~3,210‐3,230 cm 1, indicating a strong polarization effect (Figures 3b and S6). We note that the orientation of a single‐crystal platelet could affect not only absorbance intensities but also band positions, resulting in offsets of peak positions and intensities between platelets (Jacobsen et al., 2005). We should note that in brucite two OH‐stretching bands were found − − to be sharp with FWHM less than 50 cm 1 and centered at 3,240 and 3,698 cm 1 (Frost & − Kloprogge, 1999), while in super hydrous phase B, two IR bands at 3,347 and 3,407 cm 1 or one rela- − tively broader band at 3,373 cm 1 were observed for low‐temperature ordered or high‐temperature dis- ordered polymorphs, respectively (Koch‐Müller et al., 2005). The IR bands from brucite and super

FU ET AL. 6 Geophysical Research Letters 10.1029/2019GL084630

Figure 3. FTIR and NanoSIMS results on two single‐crystal (Al,Fe)‐bearing bridgmanite platelets with crystallographic orientations of (−0.50, 0.05, −0.88) (P1) and (0.16, 0.13, −0.98) (P2). (a) Representative unpolarized IR spectra of P1 − and P2 platelets. Two IR absorption bands are observed for both platelets, centering at ~3,210 and ~3,465 cm 1 for P1 and − at ~3,230 and ~3,475 cm 1 for P2. Thickness of both platelets is ~20 μm. (b) Representative polarized IR spectra of P1 platelet at different azimuthal angles. The angles represent the rotation of the crystal platelet about its normal axis. − Numbers in (a) and (b) next to peaks show OH‐stretching frequencies with a unit of cm 1. (c and d) Optical images of P1 and P2 platelets in Ir holders under reﬂected light. The images were taken right before NanoSIMS analyses. Three to four areas labeled as red dots were selected for NanoSIMS measurements. Numbers next to red dots represent determined water concentrations in ppm wt.

hydrous phase B are not observed in measured FTIR spectra, indicating no such hydrous inclusions in our bridgmanite crystals within the spectral resolution of FTIR measurements. Reliable calibrations for absorption coefﬁcients and orientation factors of bridgmanite crystals in FTIR spectra analyses have yet to be established at this point in time. This has prevented us and all the other previous investigators to reliably evaluate the water concentrations in bridgmanite with or without Fe/Al substitution. In order to roughly estimate the hydrogen concentration from FTIR spectra, we followed the calibration method of Paterson (1982). After appropriate subtraction of baselines, the water content could be calculated using the equation:

X kðÞν C ¼ i ∫ dν OH 150ξ ðÞ3780−ν

6 where COH has a unit of ppm wt water; Xi is the density factor in cm ppm wt water, calculated as Xi =10 × (18/2d) with d as the mineral density of 4,200 g/L, and Xi is calculated to be 2,142 cm ppm wt water for the − (Al,Fe)‐bearing bridgmanite used here; kðÞν is the absorption coefﬁcient in cm 1 for a given wavenumber (ν) as shown in Figure 3a; ξ is the orientation factor that is assumed as 1/3 for unpolarized spectra. The value of 1/3 for ξ is taken from literatures on relatively better investigated upper‐mantle minerals (e.g., Jacobsen et al., 2005) under the assumption that mineral grains were randomly oriented under unpolarized light.

FU ET AL. 7 Geophysical Research Letters 10.1029/2019GL084630

Figure 4. Comparison of water concentrations in bridgmanite with different Fe and Al contents from this study and Litasov et al. (2003). Data sets in literatures (Bolfan‐Casanova et al., 2003; Inoue et al., 2010; Murakami et al., 2002; Panero et al., 2015) are tabulated in Table S2 but excluded in the plot because of possible contaminations and/or inﬂuences by coexisting phase(s). (a) Water concentrations as a function of Fe content (mol%) in bridgmanite; (b) Water concentrations as a function of Al content (mol%) in bridgmanite; (c) Water concentrations as a function of average Fe and Al contents (mol%) in bridgmanite, calculated as (Fe + Al)/(Mg + Fe + Si + Al). Circles are data from FTIR measurements and squares are results from SIMS/NanoSIMS measurements. Symbols are used to distinguish the effects of Fe and Al in bridgmanite: (a) Open and solid symbols are Al‐free and Al‐bearing bridgmanite, respectively; (b) open and solid symbols are Fe‐free and Fe‐bearing bridgmanite, respectively; and (c) open and solid symbols are endmember MgSiO3 and (Al,Fe)‐bearing bridgmanite, respectively. FTIR = Fourier‐transform infrared; SIMS = secondary ion mass spectrometry.

Under these conditions, water concentrations in P1 and P2 platelets were calculated as 731(±68) and 596(±53) ppm wt, respectively (Table S2). NanoSIMS measurements show the average water concentration is about 1,020(±70) ppm wt for both platelets (Figures 3c and 3d). The variations of measured water concentration in different spots of the platelets are typically less than 70 ppm wt at ±2σ level. All the other volatile elements (C, S, F, and Cl) are less than 1 ppm wt, indicating minimal contaminations from epoxy or residual hydrocarbons (e.g., acetone) on the surface. Our results show that the measured water concentrations from NanoSIMS are similar but greater than the estimated values from FTIR. Literature studies indicate that SIMS analyses commonly result in higher water concentrations in minerals than FTIR (e.g., Mosenfelder et al., 2015). Bolfan‐Casanova et al. (2003) stated a possible reason that SIMS cannot discriminate hydrogen counts from microscopic inclusions within samples. Such contaminations cannot be ruled out in previous studies (Bolfan‐Casanova et al., 2000; Bolfan‐ Casanova et al., 2003; Litasov et al., 2003; Murakami et al., 2002; Panero et al., 2015). We note that the collected HRTEM images on our bridgmanite crystals can be used to exclude the possibility that precipitation of water or hydrous inclusions were captured within the crystal in nanometer‐scale resolution. Thus, we argue that such inconsistency between FTIR and NanoSIMS could come from the FTIR data analysis procedure, where ξ was simply assumed as 1/3 due to the lack of such information. Early studies indicate that ξ can be affected by many factors including the nature of spectra, thickness of the sample, and degree of polarization anisotropy (Bell et al., 2003). They also show that the use of calibration method by Paterson (1982) to analyze unpolarized IR data on oriented olivine crystal platelets would underestimate its water concentration by 2.3–3.5 times due to the choice of ξ (Bell et al., 2003). Due to the lack of reliable calibration of ξ for bridgmanite, we believe the roughly calculated 600–730 ppm wt water from FTIR analyses should be an underestimate for the water concentration in our bridgmanite crystals. Considering the well‐characterized high‐quality nature of bridgmanite crystals used here, the results from NanoSIMS should represent a reliable characterization of its lattice‐bonded water. The strong OH‐stretching bands in FTIR spectra provide sup- portive evidence of its high water concentration. A comparison between previous experimental observations (e.g., Litasov et al., 2003; Meade et al., 1994) and this study shows that, for the ﬁrst order, water concentrations in bridgmanite increase with increasing Fe and/or Al contents (Figure 4). We note that some data sets in the literatures (Bolfan‐Casanova et al., 2003; Inoue et al., 2010; Murakami et al., 2002; Panero et al., 2015) are excluded in the comparison because water contents in these studies could be contaminated by inclusions and/or coexisting phases. We speculate that hydrogen substitution in bridgmanite involves a link to Fe and/or Al substitution. Two hydrogen substitution mechanisms have been proposed to link the high water concentrations in Al‐bearing bridgmanite

FU ET AL. 8 Geophysical Research Letters 10.1029/2019GL084630

(Litasov et al., 2003): (1) cation vacancy associated with Al3+ substitution at the Si site via Si4+↔Al3++H+ 4+ 3+ (e.g., Hernández et al., 2013) and (2) interstitial sites associated with oxygen vacancies 2Si ↔2Al +VO, 2‐ ‐ which can be hydrated via VO+O +H2O↔2OH (Litasov et al., 2003; Navrotsky, 1999). Both mechanisms result in Al3+ in the Si site, which attracts protons and could explain the enhanced hydrogen solubility in Al‐bearing bridgmanite. In either case, there will be a volume expansion. With the limited data presented here, it is difﬁcult to constrain hydrogen substitution mechanism in our bridgmanite crystals at an atomistic level, but the peak frequencies in FTIR spectra represent the strength of OH‐stretching bands that could help us identify the relevant location of hydrogen (e.g., Balan et al., 2008; Jacobsen et al., 2005; Tollan et al., 2017). Previous studies on olivine (e.g., Balan et al., 2008; Tollan et al., 2017) provide a hint for assigning the two absorption peaks observed here to possible mechanisms: − the peak at ~3,460 cm 1 might correspond to hydrogen substitution associated with Al3+ (Si site) and the − other one at ~3,230 cm 1 might correspond to hydrogen trapped in the Mg‐site vacancy (Table S2; e.g., Balan et al., 2008; Tollan et al., 2017). The former is consistent with the model of Al3+ incorporation in the Si site in bridgmanite (e.g., Hernández et al., 2013; Litasov et al., 2003; Navrotsky, 1999), while the latter is consistent with Fe3+ mainly substituting in the Mg site, as revealed by our Mössbauer spectroscopy results. This substitution would promote the formation of Mg‐site vacancies.

4. Geophysical Implication Bridgmanite and ferropericlase are the two most abundant minerals in the Earth's lower mantle, comprising of about 75 and 18 vol% of the region, respectively (Irifune et al., 2010; Ringwood, 1975). Extensive experiments (Bolfan‐Casanova et al., 2002; Bolfan‐Casanova et al., 2006; Joachim et al., 2013; Litasov, 2010) and ab initio calculations (Merli et al., 2016; Muir & Brodholt, 2018) show that the water concentration in ferropericlase is rather limited within the range of 10–80 ppm wt. These studies also suggest that water in the lower mantle prefers to exist in (Al,Fe)‐bearing bridgmanite. Taking into account deep‐mantle compositional models (Irifune et al., 2010; Ringwood, 1975), our (Al,Fe)‐bearing bridgmanite could be representative of the Fe, Al, and Fe3+/ΣFe (~0.65) contents in the lower mantle (Frost et al., 2004; McCammon, 1997). The high water concentration, ~1,020(±70) ppm wt, suggests that in comparison to ferropericlase, (Al,Fe)‐bearing bridgmanite should be the main water carrier in the lower mantle. The high water concentration in lower‐mantle bridgmanite could greatly affect its chemical and physical properties. The most direct and important effect is likely the influence on melting. Seismic low‐velocity observations at the topmost lower mantle (below 660‐km discontinuity) has been attributed to dehydration melting of hydrous ringwoodite transitioning to bridgmanite (e.g., Liu et al., 2016; Liu et al., 2018; Schmandt et al., 2014). Previous studies on water content in the transition zone have been controversial, but an estimate from an electrical conductivity study (Karato, 2011) shows ~0.1 wt% on average with large regional variations. Our findings show that ~0.1 wt% water could be dissolved into bridgmanite (the water content in solidus) with the coexisting melt containing ~13–18 wt% water (the water content in liquidus; the range reflects possible uncertainties in the estimate) at approximately 24 GPa and 1800 °C (see Text S3 in supporting information). The P‐T conditions at 660‐km depth are expected to be ~23.6 GPa and ~1600 °C (Katsura et al., 2010), where bridgmanite would likely be able to accommodate slightly more water than what we observed here. These observations indicate that for dehydration melting to occur at the topmost lower mantle near the 660‐km depth, the water concentration in downward flow of transition zone materials must exceed the solubility limit of about 0.1 wt%. When dehydration melting occurs, the melt will contain a large amount of water at the topmost lower mantle and hence likely affect the characteristics of seismic waves and contribute to the nature of global water circulation (Liu et al., 2016; Liu et al., 2018). We note that the bridgmanite crystals in this study were synthesized at relatively shallow lower‐mantle P‐T conditions. To understand water concentration and possible dehydration mechanism in deeper parts of lower mantle, extrapolation to higher P‐T conditions needs to be taken into account but with great care. Hydrogen dissolved into bridgmanite might affect its plastic deformation (diffusion) and electrical conductivity. However, the magnitude of hydrogen substitution effects on these properties for bridgmanite is poorly constrained at this time. High P‐T experimental studies on the effects of hydrogen in bridgmanite are crucially needed to better understand the dynamic process of Earth's lower mantle.

FU ET AL. 9 Geophysical Research Letters 10.1029/2019GL084630

Acknowledgments References The authors thank Z. Jiang for his Asimow, P. D., Stein, L. C., Mosenfelder, J. L., & Rossman, G. R. (2006). Quantitative polarized infrared analysis of trace OH in populations assistance with FTIR measurements at of randomly oriented mineral grains. American Mineralogist, 91(2‐3), 278–284. https://doi.org/10.2138/am.2006.1937 Yale University and a discussion on Aubaud, C., Withers, A. C., Hirschmann, M. M., Guan, Y., Leshin, L. A., Mackwell, S. J., & Bell, D. R. (2007). Intercalibration of FTIR and TEM results. The authors thank C. SIMS for hydrogen measurements in glasses and nominally anhydrous minerals. American Mineralogist, 92(5‐6), 811–828. https://doi. McCammon for helping with org/10.2138/am.2007.2248 Mössbauer spectroscopy data collection Balan, E., Refson, K., Blanchard, M., Delattre, S., Lazzeri, M., Ingrin, J., et al. (2008). Theoretical infrared absorption coefficient of OH and data analysis on the bridgmanite groups in minerals. American Mineralogist, 93(5‐6), 950–953. https://doi.org/10.2138/am.2008.2889 crystals. We acknowledge V. Bell, D. R., Rossman, G. R., Maldener, J., Endisch, D., & Rauch, F. (2003). Hydroxide in olivine: A quantitative determination of the Prakapenka for the assistance with X‐ absolute amount and calibration of the IR spectrum. Journal of Geophysical Research, 108(B2), 2105. https://doi.org/10.1029/ ray diffraction experiments at 13ID‐D, 2001JB000679 GSECARS. GSECARS operations are Bercovici, D., & Karato, S.‐i. (2003). Whole‐mantle convection and the transition‐zone water filter. Nature, 425(6953), 39–44. https://doi. supported by the National Science org/10.1038/nature01918 Foundation‐Earth Sciences (EAR‐ Bolfan‐Casanova, N. (2005). Water in the Earth's mantle. Mineralogical Magazine, 69(3), 229–257. https://doi.org/10.1180/ 1128799) and the U.S. Department of 0026461056930248 Energy, Geosciences (DE‐FG02‐ ‐ – – 94ER14466). J. F. L. acknowledges Bolfan Casanova, N., Keppler, H., & Rubie, D. C. (2000). Water partitioning between nominally anhydrous minerals in the MgO SiO2 ‐ support from National Science H2O system up to 24 GPa: Implications for the distribution of water in the Earth's mantle. Earth and Planetary Science Letters, 182(3 4), – ‐ ‐ Foundation Geophysics Program (EAR‐ 209 221. https://doi.org/10.1016/S0012 821X(00)00244 2 ‐ 1446946 & EAR‐1916941) and Deep Bolfan Casanova, N., Keppler, H., & Rubie, D. C. (2003). Water partitioning at 660 km depth and evidence for very low water solubility in Carbon Observatory of the Sloan magnesium silicate perovskite. Geophysical Research Letters, 30(17), 1905. https://doi.org/10.1029/2003GL017182 ‐ Foundation. S. K. acknowledges Bolfan Casanova, N., Mackwell, S., Keppler, H., McCammon, C., & Rubie, D. (2002). Pressure dependence of H solubility in support from EAR‐1082622. A. G. G. magnesiowüstite up to 25 GPa: Implications for the storage of water in the Earth's lower mantle. Geophysical Research Letters, 29(10), acknowledges support of RScF 16‐12‐ 1449. https://doi.org/10.1029/2001GL014457 ‐ 10464 grant and Center for Collective Bolfan Casanova, N., McCammon, C. A., & Mackwell, S. J. (2006). Water in transition zone and lower mantle minerals. In Earth's Deep – Use “Accelerator Center for Neutron Water Cycle, (pp. 57 68). Washington, DC: American Geophysical Union. ‐ Research of the Structure of Substance Dai, L., & Karato, S. i. (2009). Electrical conductivity of wadsleyite at high temperatures and high pressures. Earth and Planetary Science ‐ – and Nuclear Medicine” of the INR RAS. Letters, 287(1 2), 277 283. https://doi.org/10.1016/j.epsl.2009.08.012 ‐ ‐ TEM analyses were supported by the Dera, P., Zhuravlev, K., Prakapenka, V., Rivers, M. L., Finkelstein, G. J., Grubor Urosevic, O., et al. (2013). High pressure single crystal ‐ – Ministry of Science and Higher micro X ray diffraction analysis with GSE_ADA/RSV software. High Pressure Research, 33(3), 466 484. https://doi.org/10.1080/ Education within the State assignment 08957959.2013.806504 FSRC «Crystallography and Photonics» Duffy, T. S., Meade, C., Fei, Y., Mao, H.‐K., & Hemley, R. J. (1995). High‐pressure phase transition in brucite, Mg (OH) 2. American RAS. Experimental data for EPMA are Mineralogist, 80(3‐4), 222–230. https://doi.org/10.2138/am‐1995‐3‐403 ‐ listed in Table S1. Raw unpolarized and Durben, D. J., & Wolf, G. H. (1992). High temperature behavior of metastable MgSiO3 perovskite: A Raman spectroscopic study. American polarized FTIR data are available in the Mineralogist, 77(7‐8), 890–893. supporting information. All the data to Frost, D. J., Liebske, C., Langenhorst, F., McCammon, C. A., Trønnes, R. G., & Rubie, D. C. (2004). Experimental evidence for the existence produce all the figures in this paper are of iron‐rich metal in the Earth's lower mantle. Nature, 428(6981), 409–412. https://doi.org/10.1038/nature02413 available on Zenodo (http://doi.org/ Frost, R. L., & Kloprogge, J. T. (1999). Infrared emission spectroscopic study of brucite. Spectrochimica Acta Part A: Molecular and 10.5281/zenodo.3364107). More Biomolecular Spectroscopy, 55(11), 2195–2205. https://doi.org/10.1016/S1386‐1425(99)00016‐5 detailed information of experimental Hauri, E. H., Weinreich, T., Saal, A. E., Rutherford, M. C., & van Orman, J. A. (2011). High pre‐eruptive water contents preserved in lunar results can be found in the supporting melt inclusions. Science, 333(6039), 213–215. https://doi.org/10.1126/science.1204626 information. Hemley, R., Cohen, R., Yeganeh‐Haeri, A., Mao, H., Weidner, D., & Ito, E. (1989). Raman spectroscopy and lattice dynamics of MgSiO3‐ perovskite at high pressure. Perovskite: A Structure of Great Interest to Geophysics and Materials Science, 45,35–44. Hernández, E. R., Alfe, D., & Brodholt, J. (2013). The incorporation of water into lower‐mantle perovskites: A first‐principles study. Earth and Planetary Science Letters, 364,37–43. https://doi.org/10.1016/j.epsl.2013.01.005 Hofmeister, A., Cynn, H., Burnley, P., & Meade, C. (1999). Vibrational spectra of dense, hydrous magnesium silicates at high pressure: Importance of the hydrogen bond angle. American Mineralogist, 84(3), 454–464. https://doi.org/10.2138/am‐1999‐0330

Inoue, T. (1994). Effect of water on melting phase relations and melt composition in the system Mg2SiO4 ▪ MgSiO3 ▪ H2O up to 15 GPa. Physics of the Earth and Planetary Interiors, 85(3‐4), 237–263. https://doi.org/10.1016/0031‐9201(94)90116‐3 Inoue, T., Wada, T., Sasaki, R., & Yurimoto, H. (2010). Water partitioning in the Earth's mantle. Physics of the Earth and Planetary Interiors, 183(1‐2), 245–251. https://doi.org/10.1016/j.pepi.2010.08.003 Irifune, T., Shinmei, T., McCammon, C. A., Miyajima, N., Rubie, D. C., & Frost, D. J. (2010). Iron partitioning and density changes of pyrolite in Earth's lower mantle. Science, 327(5962), 193–195. https://doi.org/10.1126/science.1181443 Jacobsen, S. D., Demouchy, S., Frost, D. J., Ballaran, T. B., & Kung, J. (2005). A systematic study of OH in hydrous wadsleyite from polarized FTIR spectroscopy and single‐crystal X‐ray diffraction: Oxygen sites for hydrogen storage in Earth's interior. American Mineralogist, 90(1), 61–70. https://doi.org/10.2138/am.2005.1624 Joachim, B., Wohlers, A., Norberg, N., Gardés, E., Petrishcheva, E., & Abart, R. (2013). Diffusion and solubility of hydrogen and water in periclase. Physics and Chemistry of Minerals, 40(1), 19–27. https://doi.org/10.1007/s00269‐012‐0542‐8 Kaminsky, F. V. (2018). Water in the Earth's lower mantle. Geochemistry International, 56(12), 1117–1134. https://doi.org/10.1134/ S0016702918120042 Karato, S. (2015). Water in the evolution of the Earth and other terrestrial planets. Treatise on Geophysics, 9, 105–144. Karato, S.‐i. (2011). Water distribution across the mantle transition zone and its implications for global material circulation. Earth and Planetary Science Letters, 301(3‐4), 413–423. https://doi.org/10.1016/j.epsl.2010.11.038 Karato, S.‐I., & Jung, H. (2003). Effects of pressure on high‐temperature dislocation creep in olivine. Philosophical Magazine, 83(3), 401–414. https://doi.org/10.1080/0141861021000025829 Karato, S.‐i., & Wang, D. (2013). Electrical conductivity of minerals and rocks. Physics and Chemistry of the Deep Earth, 145‐182. https://doi. org/10.1002/9781118529492.ch5 Katsura, T., Yoneda, A., Yamazaki, D., Yoshino, T., & Ito, E. (2010). Adiabatic temperature proﬁle in the mantle. Physics of the Earth and Planetary Interiors, 183(1‐2), 212–218. https://doi.org/10.1016/j.pepi.2010.07.001

Kawamoto, T. (2004). Hydrous phase stability and partial melt chemistry in H2O‐saturated KLB‐1 peridotite up to the uppermost lower mantle conditions. Physics of the Earth and Planetary Interiors, 143, 387–395.

FU ET AL. 10 Geophysical Research Letters 10.1029/2019GL084630

Koch‐Müller, M., Dera, P., Fei, Y., Hellwig, H., Liu, Z., van Orman, J., & Wirth, R. (2005). Polymorphic phase transition in superhydrous phase B. Physics and Chemistry of Minerals, 32(5‐6), 349–361. https://doi.org/10.1007/s00269‐005‐0007‐4 Kojitani, H., Katsura, T., & Akaogi, M. (2007). Aluminum substitution mechanisms in perovskite‐type MgSiO3: An investigation by Rietveld analysis. Physics and Chemistry of Minerals, 34(4), 257–267. https://doi.org/10.1007/s00269‐007‐0144‐z Kumamoto, K. M., Warren, J. M., & Hauri, E. H. (2017). New SIMS reference materials for measuring water in upper mantle minerals. American Mineralogist, 102(3), 537–547. https://doi.org/10.2138/am‐2017‐5863CCBYNCND Litasov, K. (2010). The inﬂuence of Al2O3 on the H2O content in periclase and ferropericlase at 25 GPa. Russian Geology and Geophysics, 51(6), 644–649. https://doi.org/10.1016/j.rgg.2010.05.005 Litasov, K., Ohtani, E., Langenhorst, F., Yurimoto, H., Kubo, T., & Kondo, T. (2003). Water solubility in Mg‐perovskites and water storage capacity in the lower mantle. Earth and Planetary Science Letters, 211(1‐2), 189–203. https://doi.org/10.1016/S0012‐ 821X(03)00200‐0 Litasov, K. D., Kagi, H., Shatskiy, A., Ohtani, E., Lakshtanov, D. L., Bass, J. D., & Ito, E. (2007). High hydrogen solubility in Al‐rich stishovite and water transport in the lower mantle. Earth and Planetary Science Letters, 262(3‐4), 620–634. https://doi.org/10.1016/j. epsl.2007.08.015 Liu, Z., Park, J., & Karato, S. I. (2016). Seismological detection of low‐velocity anomalies surrounding the mantle transition zone in Japan subduction zone. Geophysical Research Letters, 43, 2480–2487. https://doi.org/10.1002/2015GL067097 Liu, Z., Park, J., & Karato, S.‐i. (2018). Seismic evidence for water transport out of the mantle transition zone beneath the European Alps. Earth and Planetary Science Letters, 482,93–104. https://doi.org/10.1016/j.epsl.2017.10.054 Lloyd, A. S., Ruprecht, P., Hauri, E. H., Rose, W., Gonnermann, H. M., & Plank, T. (2014). NanoSIMS results from olivine‐hosted melt embayments: Magma ascent rate during explosive basaltic eruptions. Journal of Volcanology and Geothermal Research, 283,1–18. https://doi.org/10.1016/j.jvolgeores.2014.06.002 McCammon, C. (1997). Perovskite as a possible sink for ferric iron in the lower mantle. Nature, 387(6634), 694–696. https://doi.org/ 10.1038/42685 Meade, C., Reffner, J. A., & Ito, E. (1994). Synchrotron infrared absorbance measurements of hydrogen in MgSiO3 perovskite. Science, 264(5165), 1558–1560. https://doi.org/10.1126/science.264.5165.1558 Mei, S., & Kohlstedt, D. (2000). Inﬂuence of water on plastic deformation of olivine aggregates: 1. Diffusion creep regime. Journal of Geophysical Research, 105(B9), 21,457–21,469. https://doi.org/10.1029/2000JB900179 Merli, M., Bonadiman, C., Diella, V., & Pavese, A. (2016). Lower mantle hydrogen partitioning between periclase and perovskite: A quantum chemical modelling. Geochimica et Cosmochimica Acta, 173, 304–318. https://doi.org/10.1016/j.gca.2015.10.025 Mosenfelder, J. L., le Voyer, M., Rossman, G. R., Guan, Y., Bell, D. R., Asimow, P. D., & Eiler, J. M. (2011). Analysis of hydrogen in olivine by SIMS: Evaluation of standards and protocol. American Mineralogist, 96(11‐12), 1725–1741. https://doi.org/10.2138/ am.2011.3810 Mosenfelder, J. L., Rossman, G. R., & Johnson, E. A. (2015). Hydrous species in feldspars: A reassessment based on FTIR and SIMS. American Mineralogist, 100(5‐6), 1209–1221. https://doi.org/10.2138/am‐2015‐5034 Muir, J. M., & Brodholt, J. P. (2018). Water distribution in the lower mantle: Implications for hydrolytic weakening. Earth and Planetary Science Letters, 484, 363–369. https://doi.org/10.1016/j.epsl.2017.11.051 Murakami, M., Hirose, K., Yurimoto, H., Nakashima, S., & Takafuji, N. (2002). Water in Earth's lower mantle. Science, 295(5561), 1885–1887. https://doi.org/10.1126/science.1065998 Navrotsky, A. (1999). A lesson from ceramics. Science, 284(5421), 1788–1789. https://doi.org/10.1126/science.284.5421.1788 Ohtani, E. (2015). Hydrous minerals and the storage of water in the deep mantle. Chemical Geology, 418,6–15. https://doi.org/10.1016/j. chemgeo.2015.05.005 Ohtani, E., Toma, M., Litasov, K., Kubo, T., & Suzuki, A. (2001). Stability of dense hydrous magnesium silicate phases and water storage capacity in the transition zone and lower mantle. Physics of the Earth and Planetary Interiors, 124(1‐2), 105–117. https://doi.org/10.1016/ S0031‐9201(01)00192‐3 Okuchi, T., Purevjav, N., Tomioka, N., Lin, J.‐F., Kuribayashi, T., Schoneveld, L., et al. (2015). Synthesis of large and homogeneous single crystals of water‐bearing minerals by slow cooling at deep‐mantle pressures. American Mineralogist, 100(7), 1483–1492. https://doi.org/ 10.2138/am‐2015‐5237 Panero, W. R., Pigott, J. S., Reaman, D. M., Kabbes, J. E., & Liu, Z. (2015). Dry (Mg, Fe) SiO3 perovskite in the Earth's lower mantle. Journal of Geophysical Research: Solid Earth, 120, 894–908. https://doi.org/10.1002/2014JB011397 Paterson, M. (1982). The determination of hydroxyl by infrared absorption in quartz, silicate glasses and similar materials. Bulletin de Mineralogie, 105,20–29. Pearson, D., Brenker, F., Nestola, F., McNeill, J., Nasdala, L., Hutchison, M., et al. (2014). Hydrous mantle transition zone indicated by ringwoodite included within diamond. Nature, 507(7491), 221–224. https://doi.org/10.1038/nature13080 Peslier, A. H., Schönbächler, M., Busemann, H., & Karato, S.‐I. (2017). Water in the Earth's interior: Distribution and origin. Space Science Reviews, 212(1‐2), 743–810. https://doi.org/10.1007/s11214‐017‐0387‐z A. E. Ringwood (1975), Composition and petrology of the Earth's mantle [by] AE Ringwood. Sambridge, M., Gerald, J. F., Kovács, I., O'Neill, H. S. C., & Hermann, J. R. (2008). Quantitative absorbance spectroscopy with unpolarized light: Part I. Physical and mathematical development. American Mineralogist, 93(5‐6), 751–764. https://doi.org/10.2138/ am.2008.2657 Schmandt, B., Jacobsen, S. D., Becker, T. W., Liu, Z., & Dueker, K. G. (2014). Dehydration melting at the top of the lower mantle. Science, 344(6189), 1265–1268. https://doi.org/10.1126/science.1253358 Schwager, B., Chudinovskikh, L., Gavriliuk, A., & Boehler, R. (2004). Melting curve of H2O to 90 GPa measured in a laser‐heated diamond cell. Journal of Physics: Condensed Matter, 16(14), S1177. Shatskiy, A., Katsura, T., Litasov, K., Shcherbakova, A., Borzdov, Y., Yamazaki, D., et al. (2011). High pressure generation using scaled‐up Kawai‐cell. Physics of the Earth and Planetary Interiors, 189(1‐2), 92–108. https://doi.org/10.1016/j.pepi.2011.08.001 Shatskiy, A., Litasov, K. D., Matsuzaki, T., Shinoda, K., Yamazaki, D., Yoneda, A., et al. (2009). Single crystal growth of wadsleyite. American Mineralogist, 94(8‐9), 1130–1136. https://doi.org/10.2138/am.2009.3150 Spektor, K., Nylen, J., Stoyanov, E., Navrotsky, A., Hervig, R. L., Leinenweber, K., et al. (2011). Ultrahydrous stishovite from high‐pressure hydrothermal treatment of SiO2. Proceedings of the National Academy of Sciences, 108(52), 20,918–20,922. https://doi.org/10.1073/ pnas.1117152108 Stadelmann, P. (1987). EMS‐a software package for electron diffraction analysis and HREM image simulation in materials science. Ultramicroscopy, 21(2), 131–145. https://doi.org/10.1016/0304‐3991(87)90080‐5

FU ET AL. 11 Geophysical Research Letters 10.1029/2019GL084630

Tauzin, B., Debayle, E., & Wittlinger, G. (2010). Seismic evidence for a global low‐velocity layer within the Earth's upper mantle. Nature Geoscience, 3(10), 718–721. https://doi.org/10.1038/ngeo969 Tollan, P. M., Smith, R., O'Neill, H. S. C., & Hermann, J. (2017). The responses of the four main substitution mechanisms of H in olivine to H2O activity at 1050 °C and 3 GPa. Progress in Earth and Planetary Science, 4(1), 14. https://doi.org/10.1186/s40645‐017‐0128‐7

FU ET AL. 12